ATime SeriesAnalysisOfCrimeTrinidadandTobago

8/13/2019 ATime SeriesAnalysisOfCrimeTrinidadandTobago

http://slidepdf.com/reader/full/atime-seriesanalysisofcrimetrinidadandtobago 1/16

A Time-Series Analysis of Crime in Trinidad and Tobago

Sandra Sookram*

Maukesh Basdeo**

Kerry Sumesar-Rai**

George Saridakis***

* Sir Arthur Lewis Institute of Social and Economic StudiesThe University of the West Indies

**Department of Behavioural Sciences

The University of the West Indies

*** The Business School

Loughborough University

SALISES Publications • Working Papers

Paper 2009:20



Abstract

What should be done about crime is a matter of heated debate: popular, political, and

academic. Alternatively, one might treat crime control as an ordinary policy issue, like

transportation or the control of pollution, in which policies are to be judged by their

outcomes, including their costs. In Trinidad and Tobago the policy agenda and the public are

demanding stronger measures to deal with rising crime. This paper evaluates empirically the

effect of the criminal justice system and socio-economic conditions on serious crime (which

includes both property and violent crime) in Trinidad and Tobago over the period 1970-2007.

Our study finds that the crime detection rate, unemployment rate, the percentage of females

in the labour force and the percentage of the labour force with tertiary education are very

important in determining criminal behaviour.

Keywords: Crime, Trinidad and Tobago, Time-Series

JEL classification: I2, K42



1. Introduction

The 2007 World Bank report revealed that murder rates in the Caribbean region is

higher than in any other part of the world. In particular, Trinidad and Tobago (T&T),

a unitary state in the Caribbean region, that has seen its overall crime rate escalating at

a phenomenal rate especially within the last decade. At the same time, crime detection

has been experiencing a significant drop (it accounts to about 42% for serious crime)

which may explain the dramatic increase in crime and delinquency in the 2000s. More

recently Mohammed et al. (2009) indicate that past victimization, especially when no

police action was taken to arrest the perpetrator, perpetuated and increased the

perceived fear of crime in T&T. Nevertheless, research on criminal activity in

Trinidad and Tobago (and the Caribbean) has been generally scarce.

This is therefore a timely paper that aims to examine the effect of police crime

detection measured by the proportion of crimes which are cleared up and various

socio-economic factors on crime rates in T&T over the last four decades. Serious

crime and some specific crimes are examined in this literature. Our modelling strategy

is based mainly on the economic crime literature, which developed following Becker

(1968) and Ehrlich (1973) path-breaking work, and to a lesser extent references to

criminological literature are also made. Our empirical methodology is based on

Johansen’s (1995 and 1998) procedure, which is the most commonly used system

method in cointegration analysis.

The rest of the paper is organised as follows: in section 2 we examine the main

characteristics of the crime rate in Trinidad and Tobago. Following this, Section 3

presents the data and an examination of the independent variables in the model.

Section 4 presents and discusses the results. The conclusions are presented in Section

5.



2. Crime in Trinidad and Tobago: Some Stylized Details

Data collected from the Crime and Problem Analysis (CAPA) Unit of the Ministry of

National Security indicate that violent crime rose rapidly in the 1980s, began

decreasing during the mid-1990s, and rose again in the 2000s. At the same time,

property crime rose significantly in the 1970s, levelled off in the 1980s and 1990s,

and increased sharply in the 2000s. In fact, property crime by 2007 has almost

quadrupled the figure for 1970. Turning to the overall crime rate, it reached its zenith

in 1988, with a dramatic rise during the 1980s. During this decade the country was in

the trough of an economic recession, but there were more minor crimes than serious

crimes1. The incidence of minor crimes is more likely to occur during periods of

economic downturns when social and political instability seem more likely to occur.

It was only in the 1970s that serious crimes out-numbered minor crimes and this

pattern began to reoccur in 2000s2. Significantly, these two decades represented

periods of economic prosperity. Figure 1 summarises criminal activity in T&T during

the last four decades.

Focusing on serious crime, we further discuss two of its more heinous

components, that is, murder and kidnapping. These two types of crime have

traumatised the society as they have increased significantly over the recent years. It is

possible that changes in government, with the accompanying changes in policy and

focus, could have impacted on the serious crime rate. For example, there were two

changes in government during the period 1995 to 20073. In 1994 143 murders were

committed, but by the end of 1995 the murder figures were reduced to 122. Over the

1Minor crime are all crimes carrying a penalty of under 5 years. Serious crime includes both violentcrime and property crime as well as other serious crimes, such as fraud offences, firearm offences and

narcotic offences.2 Although from the year 1996 signs of an economic turn around were clearly visible it was not until

the second oil boom starting in 2002 that the economy fully recovered from the past recession.3 In 1995 the opposition party attained government and in 2001 the former ruling party was once again

in government.



next four years (1996 to 1999) the murder rate steadily declined. Added to the

increase in number of murders, a new type of violent crime emerged: kidnappings and

kidnappings for ransom (see Table 1).

By the end of 2000 and the beginning of the new millennium the murder rate

began to rise again. In 2001 murders increased by 23 percent as against the 1995

figures; 2002: 40 percent; 2003: 88 percent; 2004: 113 percent; 2005: 216 percent;

2006: 202 percent and 2007: 224 percent. While in general there has been an upward

trend in the number of murders, there have been fluctuations in the number of

kidnappings.

Table 1: Number of Murders and Kidnappings: 1995-2007

Year Murders Kidnappings

1995 122

1996 106

1997 101

1998 98

1999 93

2000 118

2001 151

2002 171 292003 229 51

2004 260 28

2005 386 58

2006 368 17

2007 395 155

Source: Crime in Trinidad and Tobago. http://www.ttcrime.com/stats.php

Accessed 10th July 2008



Figure 1. Crime in T&T, 1970-2007

Violent Crime Property crime

0

500

1000

1500

2000

2500

1 9 7 0

1 9 7 3

1 9 7 6

1 9 7 9

1 9 8 2

1 9 8 5

1 9 8 8

1 9 9 1

1 9 9 4

1 9 9 7

2 0 0 0

2 0 0 3

2 0 0 6

0

2000

4000

6000

8000

10000

12000

14000

16000

1 9 7 0

1 9 7 3

1 9 7 6

1 9 7 9

1 9 8 2

1 9 8 5

1 9 8 8

1 9 9 1

1 9 9 4

1 9 9 7

2 0 0 0

Serious crime and minor crime

0

5000

10000

15000

20000

25000

30000

35000

40000

45000

50000

1 9 7 0

1 9 7 3

1 9 7 6

1 9 7 9

1 9 8 2

1 9 8 5

1 9 8 8

1 9 9 1

1 9 9 4

1 9 9 7

2 0 0 0

2 0 0 3

2 0 0 6

0% 20% 40% 60% 80%

1970

1973

1976

1979

1982

1985

1988

1991

1994

1997

2000

2003

2006

Serious Crime Minor Crime

Overall crime



Clearly the increasing number of murders and the prevalence of kidnappings

are areas for serious concern for the relevant authorities, policy makers and the

citizenry. Another area of concern seems to be the rate of detection for serious

crimes. Over the period 1997 to 2003 there were significant declines in the rate of

detection for crimes such as: Murder and Wounding and Shootings (see Table 2).

With regard to the other types of serious crime the rate of detection either slightly

Table 2. Serious Crime and Detection for Selective Years, 1997-2003

Type/ Year 1997 2000 2003

No. Det. Rate

of

Det.

(%)

No. Det. Rate

of

Det.

(%)

No. Det. Rate

of

Det.

(%)

Murder 101 77 76 118 62 52 229 92 40

Wounding andShootings

370 215 58 383 188 49 784 333 42

Rapes, Incest

and Sexual

Offences

514 367 71 516 364 70 643 468 72

Serious

Indecency

206 145 70 170 127 74 88 64 73

Kidnappings 0 0 0 0 0 0 235 96 41

Robberies 3393 719 21 4052 608 15 4590 758 16

Burglaries andBreakings

6682 1070 16 5625 644 11 4863 710 15

Larceny of

Vehicles

2686 318 12 3008 267 9 3210 335 10

Larceny of

DwellingHouses

432 61 14 385 52 13 365 80 22

Total 14384 2972 21 14257 2312 16 15007 2936 20

Source: The Government of Trinidad and Tobago, Ministry of Planning and Development, Central Statistical Office., 2008

improved, or remain stabled in light of increases in the incidence of these crimes such

as, Rape, Incest and Sexual Offences; Robberies and Larceny.

For the years 2006 and 2007 there have been some positive changes in the

numbers of the different types of crimes. Wounding and shootings; Sexual Offences;

Serious Indecency; Burglaries; Abductions and Kidnappings for Ransom were



slightly down in some instances, but the figures for Murders and Larceny of vehicles

went up. However, the levels of criminal activities and the number of serious crimes

committed seem to be much higher in 2007 than that of the baseline year of 1970. In

2007 there were 8,236 incidences of serious crimes whereas, in 1970 there were 5015.

The difference between the numbers of crimes for the two years; 3,221 represents an

increase of 64 percent. It seems, therefore, that the Trinidad and Tobago society has

become more violent over the past thirty seven years.

Table 3. Crime statistics T&T, 2006-2007

Type 2006 2007

Murders 368 395Wounding and shootings 657 645

Sexual Offences 903 759

Serious Indecency 81 66

Burglaries 4973 4851

Abductions/ Kidnappings 214 162

Kidnappings for Ransom 16 13

Larceny- motor vehicles 1496 1740

Total 8340 8236

Source: Geisha Kowlessar, “Joseph: I never said we failed.” Trinidad Guardian, 8th January 2008, and

Crime in Trinidad and Tobago. http://www.ttcrime.com/stats.php Accessed 10th July 2008

3. Data

The data used in this study was obtained from the Central Statistical Office (CSO) of

Trinidad and Tobago4. Since time-series data for the detection rate are only available

for serious crime, we restrict our analysis to this type of crime.

Finally, the independent variables used in the study are based primarily on the

economics of crime literature and to a lesser extent on theories developed by

criminologists. Specifically, the independent variables are (see Table 4 for summary

statistics for the raw data over the 37-year period):

4 The crime rates for the last two years (2005-2006) were obtained from the Crime and Problem Analysis

(CAPA) Unit of the Ministry of National Security.



i) the unemployment rate; high rates of unemployment could be taken

to influence the opportunity cost of illegal activity (Witte and Witt,

2001). Existing research suggests that higher unemployment is

associated with greater occurrence of property crime, but this

relationship turned to be insignificant or negative for violent crime

(e.g. Saridakis, 2004; Raphael and Winter-Ebmer, 2001; Entorf and

Spengler, 2000)

ii) the percentage of the labour force with tertiary education; from an

economic point of view there are two main reasons why education

could have an important role in combating and reducing crime5.

Firstly, an education increases the returns to legitimate work and

raises the opportunity cost of engaging in criminal activity. Further,

and in most cases the penalty for illegal behaviour often leads to

imprisonment and being in prison translates to wages forgone.

Secondly, education may change an individual’s preferences in

some indirect ways, which could then affect the decision to commit

a crime (see Becker and Mulligan, 1997)

iii) the employed female population; Saridakis (2004) suggested that an

increase in the employed female population increase the time spent

out of homes and contribute to low parental supervision of children

and thus, could be an associated with an increase in violent crime

rates. However, this was not supported empirically. Also, one can

further argue that a higher female labour force participation reduces

5 Conversely, the lack of education can be a cause of crime and criminal behaviour (Usher, 1997 and

Becker and Mulligan, 1997).



guardianship (see theory developed by Cohen and Felson, 1979)

which in turns may increase property crime

iv) the crime detection rate; deterrence variables (i.e. detection rate,

probability of apprehension and severity of punishment) have been

used in economic studies to analyse property crime, since in effect it

impacts on expected returns to crime. Among others commentators,

however, Saridakis (2004, 2008, 2009) found a weak effect of the

deterrence hypothesis on violent crime.

Table 4: Summary Statistics

Variable Mean Std. Dev. Minimum Maximum

Serious Crime (per 100,000);scr

.195 .041 .050 .195

Unemployment Rate;

unemp

14.06 4.25 6.2 22.3

Labour Force with Tertiary

Education (%);educate

6.79 5.19 2 21

Rate of Detection (%) ;det

25.58 4.31 17.45 33.71

Employed Female Population (%) ; female

43.28 9.19 28 54

Source: Table generated by authors using data from the Trinidad and Tobago Central Statistical Office and the Crime and

Problem Analysis (CAPA) Unit of the Ministry of National Security. 2008

4. Results

Before estimation of the cointegrated VAR model, we need to ensure a stationary

representation of the model. The Augmented Dickey Fuller (ADF) test is one of the

most frequently used tests for unit roots. The null hypothesis is that there exists a unit

root (or that the series is I(1) or the time series are non-stationary). The results of the

ADF test are presented in Table 5. It shows that the null of a unit root of the variables

at levels cannot be rejected for any one of the variables at the 5% level of



significance. However, the series become stationary after first differencing6. Finally,

given that we have only 38 observations, we use a lag length of 2 in the VAR and

treat only the crime and deterrence variables as endogenous variables (see, for

example, Saridakis, 2009). We also found that there is no evidence of serial

correlation in the each of the endogenous equations using the LM test for up to

second-order serial correlation.

Table 5. Results of the Augmented Dickey-Fuller test for Unit Roots

Variable(s) Levels First

Diff.

lscr -2.87 (0) -5.20 (1)

lunemp -1.65 (2) -2.23 (2)

leducate -0.76 (4) -5.72 (3)

ldet -2.56 (7) -4.89 (0)

lfemale -1.57 (0) -5.03 (8)

Notes: In parentheses are the lag lengths based on AIC. Critical values at the 5% level is –3.544l denotes that the variables are measured in natural logarithms.

Exogenous: Constant, Linear Trend

We then employ Johansen’s (1988, 1995) ML-framework to test for the

existence of cointegration relations. Johansen’s ML-framework uses a rank test to

identify the number of co-integrating vectors ‘r’ that can be found in the data. The

results of the cointegration analysis based on the trace of the stochastic matrix

support the rejection of the null hypothesis r=0 and indicates that there is one

cointegrating relationship (r=1). Table 6 presents the results of the cointegration test.

6 With the sole exception being the unemployment variable. However, the Phillips-Perron unit root test,

which proposes an alternative nonparametric method of controlling for serial correlation when testingfor unit root suggests that the unemployment variable becomes stationary after first differencing.



Table 6. A Cointegration Analysis of Serious Crime

Null Alternative

Trace

Statistics

95%

Quantile

r=0 r>=1 42.50* 40.37

r<=1 r>=2 13.06 20.47

*Significant at the 5% level.

The long run relationship among the variables can be summarised in the

following estimated cointegrating relation in which the coefficient of ln(scr) is

normalized to 1 (estimated standard errors appear in parentheses):

0.05trend te0.15leduca0.53female-0.65lunempdet llscr +−+−= 85.0

Our findings indicate that during the period under study the detection rate, the

unemployment rate, the percentage of females in the labour force and the percentage

of persons with tertiary education had significant impacts on serious crime.

Specifically, our results indicate that the detection rate and crime are negatively

related, as the detection decreases the crime rate increases, providing support for the

deterrence hypothesis. Furthermore, we found a positive relationship between

unemployment and serious crime; high rates of unemployment indicate lack of legal

income opportunities, and thus serve to reduce the opportunity cost of engaging in

criminal activity.

In contrast to the argument developed in the data section and vast empirical

work (e.g. Witt and Witte 2000; Kapuscinski et al. 1998), we found a negative

relationship between the employed female population and crime. It is possible that

this result was obtained because the increased employed female population in T&T,

which can be viewed as a labour market improvement, may have raised family income -

especially in the case of female-head households. Finally, the percentage of persons

with tertiary education had a negative effect on the serious crime rate. This was an

(0.08) (0.05) (0.07) (0.01)(0.15)



expected result since as noted before an individual with this level of education would

have access to legitimate earnings which increases the opportunity cost of committing

a crime. Further, persons who have an education tend to have different preferences,

some of which negatively affects the choice to engage in criminal activity.

Furthermore, we estimated that the error correction coefficient to be at -0.723

(std. err. = 0.218), which is statistically highly significant and suggests a fast speed of

convergence to equilibrium. Finally, a good test of forecast performance is to examine

whether the VAR model predicts the turning points of the crime movements. Figure 2

plots in-sample fitted values and out of sample forecasts and suggests that the size of

forecast errors and the in-sample residuals are very similar.

Figure 2. Multivariate dynamic forecasts for the level of serious crime

LSCR

Forecast

-1.5

-2.0

-2.5

-3.0

-1.0

1972 1977 1982 1987 1992 1997 2002 2007

5. Conclusions

In this study we have empirically examined the impact of the crime detection rate, the

unemployment rate, the percentage of females in the labour force and the percentage

of the labour force with tertiary education on serious crime in T&T over the last four

decades. Our study concluded that all of these variables were very important in

determining the crime rate. Particularly, we found that the detection rate and crime are

negatively related. It is therefore likely that part of the increase in serious crime in the



2000s is due to the significant drop in arrest rate. In accordance with the existing

crime theory, we also found that unemployment is an important cause of crime

whereas education exerts a negative effect. Interesting, however, we found that rising

female employment is associated with reduction in crime. A plausible explanation

may be that the increase in the employed female population may have raised family

income, but this association needs to be explored further. Moreover, our data does not

distinguish between property and violent crime, in which case the above factors may

behave differently. This is an issue for further research.



References

Becker, G. (1968), Crime and Punishment: An Economic Approach, The Journal of

Political Economy, 76/2:169-217.

Becker, G and Mulligan, C. (1997), The Endogenous Determination of Time

Preference, The Quarterly Journal of Economics, 112/3:729-58.

Cohen, L. and Felson, M. (1979). Social Change and Crime Rate Trends: A RoutineActivity Approach, American Sociological Review, 44/4: 588-608.

Entorf, H. and Spengler, H. (2000). Socioeconomic and Demographic Factors of

Crime in Germany: Evidence From Panel Data of the German States, International

Review of Law and Economics, 20:75-106.

Ehrlich, I. (1973), Participation in Illegitimate Activities: A Theoretical and Empirical

Investigation, The Journal of Political Economy, 81/3:521-565.

Johansen, S. (1995), Likelihood-Based Inference in Cointegrated Vector Autoregressive Models, Oxford: Oxford University Press.

Johansen, S. (1988), Statistical Analysis of Cointegration Vectors, Journal of

Economic Dynamics and Control, 12:231-254.

Kapuscinski, C.A., J. Braithwaite and B. Chapman, (1998), “Unemployment and

Crime: Towards Resolving the Paradox”, Journal of Quantitative Criminology, 14:

215-243.

Mohammed, A.M., Saridakis, G. and Sookram, S. (2009), ''Do Victims of Crime FearCrime More? Empirical Evidence from the Survey of Living Conditions (2005) of

Trinidad and Tobago'', Sir Arthur Lewis Institute of Social and Economic Studies,The University of West Indies, SALISES Working Paper, No. 19.

Raphael, S. and Winter-Ebmer, R. (2001), Identifying the Effect of Unemployment on

Crime, Journal of Law and Economics, XLIV:259-283.

Saridakis, G. and Spengler, H. (2009), ''Crime, Deterrence and Unemployment in

Greece: A Panel Data Approach'', 2009, DIW Berlin, German Institute for Economic

Research Discussion Paper, No. 853.

Saridakis, G. (2008), ''Violent Crime and Incentives in the Long Run: Evidence from

England and Wales'', Loughborough University Business School Occasional PapersSeries, No. 2.

Saridakis, G. (2004), ''Violent Crime in the United States of America: A Time-Series

Analysis Between 1960-2000'', European Journal of Law and Economics, 18(2), pp

203-221, ISSN 0929-1261.

Usher, D. (1997), Education as a Deterrent to Crime, Canadian Journal of Economics, 30/2:367-84.



Witt, R. and Witte, A. (2000), Crime, Prison, and Female Labour Supply, Journal of

Quantitative Criminology, 16/1:69-85.

Witte and Witt (2001), Crime Causation: Economic Theories , Encyclopaedia of

Crime and Justice.

ATime SeriesAnalysisOfCrimeTrinidadandTobago

Documents

Transcript of ATime SeriesAnalysisOfCrimeTrinidadandTobago